UIUC Hexapod Project
Protobot: an insect-like walking robot.
The aim of this multidisciplinary research project is
to use insights from insect neurobiology to guide the design of adaptive
sensorimotor control systems for legged robots. We are developing neurally-inspired
control systems for insect-like robots that incorporate mechanosensory
signals from leg sensors and visual feedback from insect-like vision systems
to achieve agile locomotion over irregular terrain.
Virtual
Poster on the UIUC Hexapod Project
Look below for movie clip of robot
in action
PowerPoint
Presentation on UIUC Hexapod Project
Robot Design and Construction
Our robot has been modeled after insects
because of their ability to traverse various types of terrain with an agile
and fluid-like motion. The particular insect we have modeled our
robot after is Periplenta Americana, the american cockroach. These
insects are studied in the Entomology
department by Prof.
Delcomyn in his laboratory located in Morrill Hall (UIUC). The
robot leg dimensions were based on actual insect measurements provided
by his students.
The six legs are divided up into pairs of
different sized legs. The front leg pair is the shortest and each
pair gets progressively longer as you move toward the rear. In addition,
the front legs are attached on an angle (as opposed to an "engineer" attaching
them so they all would be perpendicular to the ground) and the angle of
attachment of the remaining leg pairs gets larger moving toward the rear.
The reason for the length and the angle differences are attributed to the
legs performing different functions during walking. All legs provide
support for the body's weight. The front legs however, also provide
for steering the insect. The middle legs 
contribute
to the raising and lowering of the insect's body. Whereas, the rear
legs provide most of the forward thrust. Another benefit of the legs
being on different angles is its influence on the resulting body length.
The insect's walking pattern moves adjacent legs towards each other.
If the legs were attached at right angles to the ground (the "engineering"
way) the legs would need to be spread apart so they would not collide.
The legs of the insect actually operate in different planes which allow
them to move toward each other without colliding. This enables the
insect and robot body to be significantly shorter in length.
Click to see movie draft with sound.
Enhanced Sensory Leg Design
Additional sensory capablilities have
been added to a mechanically improved leg. These sensors have been
based on the study of the insect leg sensors and their contribution to
walking [Delcomyn 96].
Visuomotor Behaviors
Our project studies how insect visual
systems (such as the fly and bee) utilize motion cues to extract distance
and motion information about targets and obstacles in the environment.
This image provided by the Bugscope
Project
at the Beckman
Institute.
Two small CCD cameras mounted on the
robot provide forward and lateral views of the visual field. Neurally-inspired
computer vision algorithms are used for extracting local and global motion
cues from the image.
This figure shows a horizontal cross-section
through the nervous system of the fly. Visual processing of retinal input
occurs in several stages (lamina, medulla, lobula, lobula plate),
with visuomotor control signals being relayed to the locomotor control
centers (thoracic ganglia) via the cervical connective. One goal
of our biorobotics research is to understand the nature of information
flowing between the visual processing areas and the locomotor control areas.
Click on image to enlarge.
Using visuomotor control algorithms
based on global pattern analysis of local motion cues, insects are able
to generate a variety of adaptive behaviors such a target tracking, obstacle
avoidance, postural stabilization, and active sensing (peering) for distance
estimation. We are currently implementing many of these visuomotor
behaviors on our robot.
Click on the image to see other behaviors.
Recent Publications
Cocatre-Zilgien, J. H., F. Delcomyn and J. M. Hart. 1996. Performance
of a muscle-like "leaky" pneumatic actuator powered by modulated air pulses.
J.
Robotic Sys. 13, 379-390.
Delcomyn, F., M. E. Nelson and J. H. Cocatre-Zilgien 1996. Sense
organs of insect legs and the selection of sensors for agile walking robots.
Int. J. Robotics Res. 15, 113-127.
Delcomyn, F. and Nelson, M.E. 1999 Architectures for a biomimetic
hexapod robot. Robotics and Autonomous Systems (in press).
Delcomyn, F. 1997. Insect walking. Encyclopedia of Neuroscience,
2nd edition (CD-ROM), edited by G. Adelman and B. Smith. Amsterdam: Elsevier.
Delcomyn, F. 1997. Insect models for robotics. Encyclopedia of
Neuroscience, 2nd edition (CD-ROM), edited by G. Adelman and B. Smith.
Amsterdam: Elsevier.
Ding, Z. and Nelson, M.E. 1995 A neural controller for single-leg
substrate-finding: a first step toward agile locomotion in insects and
robots. In: The Neurobiology of Computation, J.M. Bower, ed., Kluwer
Academic Press, pp. 379-384.
Lewis, M. A. and Nelson, M. E. 1998 Look Before You Leap: Peering
Behavior for Depth Perception. In: From Animals to Animats 5. Proceedings
of the Fifth International Conference on Simulation of Adaptive Behavior
,
R. Pfeifer, B. Blumberg, J-A Meyer, S. W. Wilson (eds), MIT Press, pp.
98-103.
Reichler, J. A. and F. Delcomyn. 1998. Control algorithms for biologically
inspired robots: A simulation testbed. In, R. Zobel and D. Moeller, eds,
Simulation--Past,
Present and Future. 12th European Simulation Multiconference, pp. 437-442.
Senior Personnel:
Prof.
Mark E. Nelson (Physiology, Biophysics, Neuroscience)
Prof. Narendra
Ahuja (Electrical & Computer Engineering, Artificial Intelligence)
Prof.
Fred Delcomyn (Entomology, Neuroscience)
Mr. John M. Hart (Research
Engineer)
Current Students:
Garrick Kremesec
Larry Lu
Jesse Reichler
Project Alumni:
Dr. Jan Cocatre-Zilgien
Dr. Zhimin Ding
Dr.
M. Anthony (Tony) Lewis
